Difference between MBP without nVidia and MBP with 750M - heat and battery life

[alexor]

I know that nVidia comes with the Optimus technology which is supposed to power off the nVidia chip when not in use (how about the VRAM?), but I never could find any concrete info on temperature and battery life differences between models who have only a CPU with integrated graphics, and models with a discrete nVidia + Optimus but with the nVidia switched off. IN theory, they should be equivalent

My engineering background makes me slightly skeptical that the temperature and power draw are the same between two such systems, when nVidia is "off" in the one with a discrete chip... no leakage currents? Are both the nVidia chip and VRAM completely powered off, i.e. no supply voltage supplied to either?

Anyone knows of an actual test, or at least a qualitative evaluation (e.g. fans kicking in more often at idle, worse battery life)?

Note that I'm interested in the case where the nVidia is kept "off" - if left on "dynamic" then the outcome is obvious.

Thanks in advance,
Alex.

 

[dusk007]

It is not Optimus. Optimus only exists on Windows and Linux not on OSX.
The difference in heat is miniscule in most instances. Under low load it just doesn't matter, under high load the nvidia chip makes up in architecture efficiency what it looses on having to power vram and such.
It is completely off with VRAM and all afaik. If activated the chips are powered on and the necessary data copied to VRAM. PCIe transfers a couple GB/s it fills up the VRAM in a very short time.

Battery life is better when the nvidia chip is not present simply because it cannot activate, ie. when you plug in a projector showing slides. At low load the nvidia chip adds almost 30% power draw but it has to stay active when you plug in anything or when a application even just sits in the background that requires the dGPU.
Iris Pro will just power down into the lowest power states whenever there is no actual load present, regardless of what displays are plugged in or what application runs in the background.

My old 2010 was a lot hotter with the 330M active but the 750M is very efficient when not in use. Fan heat wise it makes no difference is always stays quite until the CPU is used quite a bit. Also the Iris Pro when active seems to actually run at 55W TDP and under high usage heats up just as much. Removing the heat from one chip is so much less efficient then removing it from two chips that it pretty much makes up the difference.

When nvidia is off, I think the variance between the power draw of individual chips or the battery capacity variance is bigger than the power draw difference. It technically to anything I ever read about the subject should not require any extra power. Maybe the mux chip does but maybe that one is present anyway on both motherboards and in any case the power for that should be very small.

 

[thundersteele]

I found something for the OP. Look at 7:00 and later in this video review/comparison:
https://www.youtube.com/watch?v=s4rgL9dW7mw

The result is that the rMBP without 750M runs about 5º C cooler.

 

[qnxor]

@dusk007: Very good reply. I can also confirm that when running 3D things, using the Nvidia GPU runs cooler overall. See the tool that I wrote to test for overheating and throttling here (it plots a graph of temperature and frequency over time):

https://forums.macrumors.com/threads/1731178/

I posted 2 results there of GpuTest running the Tesseract x64 3D test, once using the Intel Iris Pro and then using the Nvidia 750M. You can see that the CPU shot to 100C and started throttling when using the Iris Pro. When using the Nvidia 750M, the CPU temps were lower and there was no throttling.

Also note that even after fans kicked in and the temperature dropped when using the Iris Pro, the max frequency of the CPU was still lower than when using the Nvidia 750M.

Using the Nvidia 750M in 3D is definitely recommended.

This is not unexpected. It's pretty basic physics. As @dusk007 said as well, cooling a single hot surface is more difficult than cooling two less hot surfaces of similar sizes.

Cheers.

p.s. As a side note, I'm pretty impressed with the cooling on this 15" rMBP Late 2013 that I'm now testing. The Dell XPS 15 (Late 2013) which has the same thickness, size Nvidia GPU but a slower CPU (4702HQ), gets much hotter than the rMBP. I measured max 47/42C on the front/back of the laptop while running X264 and GpuTest simultaneously, compared to 48/64C reported on the XPS 15 (see the NotebookCheck review). Ambient temp in my case was 25C, not exactly chilly to begin with. 64C on your lap is enough to cook yourself.